Analysing physiological electrograms

Abstract

Previous research has shown that the risk of sudden death due to cardiac arrhythmias can be predicted by observing the shape of recorded endocardial electrograms in response to pacing, and in particularly detecting certain small deflections in the recorded electrogram following early stimulation of the heart. A long standing problem has been the reliable detection of these small individual potentials because of the presence of noise in the recorded electrical signals created by other electrical equipment within a typical catheter laboratory. The solution described involves deriving a model of noise from a first portion of the electrogram in which a physiological signal is presumed to be absent, and transforming a second portion of the electrogram, presumed to contain a physiological signal, into the model of noise. The physiological signal can then be identified by identifying portions of signal within the second portion of the electrogram that do not conform to the model of noise.

The invention claimed is: 
     
       1. A method of analysing an electrogram to distinguish a physiological signal from noise; the method comprising:
 receiving a electrogram signal captured from a patient's heart, the electrogram signal including noise signals and physiological signals from the patient's heart, wherein:
 a first portion of the electrogram signal includes a noise signal absent a physiological signal, and 
 a second portion of the electrogram signal includes a noise signal combined with a physiological signal; 
 
 deriving a model of noise from the first portion of the electrogram signal by cross-correlation of the first portion of the electrogram signal with multiple templates that represent features of the physiological signal to produce a number of template correlated signals, the model of noise having limits within which a captured electrogram signal is defined as a noise signal and outside of which a captured electrogram signal is defined as a physiological signal; 
 transforming, based on the model of noise, the second portion of the electrogram signal to a trajectory vector signal; and 
 identifying the physiological signal in the second portion of the electrogram signal by identifying portions of the second portion of the electrogram signal that were transformed to portions of the trajectory vector signal that do not conform to the model of noise. 
 
     
     
       2. A method according to  claim 1  wherein the model of noise is derived from multiple first portions of the electrogram signal that include a noise signal absent a physiological signal. 
     
     
       3. A method according to  claim 1  comprising deriving a co-variance matrix from the template correlated signals for a plurality of first portions of the electrogram signal that each includes a noise signal absent a physiological signal. 
     
     
       4. A method according to  claim 3  comprising deriving a mean co-variance matrix from the co-variance matrices derived for each of a plurality of first portions of the electrogram signal that each includes a noise signal absent a physiological signal. 
     
     
       5. A method according to  claim 4  wherein the model of noise is expressed by deriving Eigenvectors and Eigen values from the mean co-variance matrix. 
     
     
       6. A method according to  claim 1  comprising correlating the second portion of the electrogram signal with multiple templates that represent features of the physiological signal to produce a set of template correlated signals. 
     
     
       7. A method according to  claim 1  comprising deriving a vector from a first time sample of each template correlated signal of the set; and further vectors from further time samples of each template correlated signal of the set. 
     
     
       8. A method according to  claim 7  comprising representing the vectors as points in the model of noise by projecting each vector onto each eigenvector thereby representing the original signal as a trajectory in the model of noise. 
     
     
       9. A method according to  claim 8  wherein a physiological signal is identified by determining points that lie outside the limits of the model of noise. 
     
     
       10. A method of analyzing a cardiac electrogram according to  claim 1 . 
     
     
       11. A method according to  claim 1  comprising reducing the trajectory vector signal to a time domain signal, and outputting the time domain signal. 
     
     
       12. Apparatus for analysing a cardiac electrogram to distinguish a physiological signal from noise; the apparatus comprising:
 a signal generator to generate a pacing signal; 
 an input electrode for applying the pacing signal to a patient's heart; 
 a receiving electrode to receive a cardiac electrogram signal from the patient's heart, the cardiac electrogram signal received based on application of the pacing signal to the patient's heart, the cardiac electrogram signal including noise signals and physiological signals from the patient's heart, wherein:
 a first portion of the cardiac electrogram signal includes a noise signal absent a physiological signal, and 
 a second portion of the cardiac electrogram signal includes a noise signal combined with a physiological signal; 
 
 a memory store in which the received cardiac electrogram signal is stored; and 
 a processor communicatively coupled to the memory store, the processor being responsive to executing computer instructions, to perform operations comprising:
 deriving a model of noise from the first portion of the cardiac electrogram signal by cross-correlation of the first portion of the electrogram signal with multiple templates that represent features of the physiological signal to produce a number of template correlated signals, the model of noise having limits within which a captured electrogram signal is defined as a noise signal and outside of which a captured electrogram signal is defined as a physiological signal; 
 
 transforming, based on the model of noise, the second portion of the electrogram signal to a trajectory vector signal; and 
 identifying the physiological signal in the second portion of the cardiac electrogram signal by identifying portions of the second portion of the cardiac electrogram signal that were transformed to portions of the trajectory vector signal that do not conform to the model of noise. 
 
     
     
       13. Apparatus according to  claim 12  wherein the processor is responsive to executing computer instructions to derive the model of noise from multiple first portions of the cardiac electrogram signal that include a noise signal absent a physiological signal. 
     
     
       14. Apparatus according to  claim 12  wherein the processor is responsive to executing computer instructions to derive a co-variance matrix from the template correlated signals for a plurality of first portions of the cardiac electrogram signal that each includes a noise signal absent a physiological signal. 
     
     
       15. Apparatus according to  claim 14  wherein the processor is responsive to executing computer instructions to derive a mean co-variance matrix from the co-variance matrices derived for each of a plurality of first portions of the cardiac electrogram signal that each includes a noise signal absent a physiological signal. 
     
     
       16. Apparatus according to  claim 15  wherein the processor is responsive to executing computer instructions to express the model by deriving Eigenvectors and Eigen values from the mean co-variance matrix. 
     
     
       17. Apparatus according to  claim 12  wherein the processor is responsive to executing computer instructions to correlate the second portion of the cardiac electrogram signal with multiple templates that represent features of the physiological signal to produce a set of template correlated signals. 
     
     
       18. Apparatus according to  claim 12  wherein the processor is responsive to executing computer instructions to derive a vector from a first time sample of each template correlated signal of the set; and further vectors from further time samples of each template correlated signal of the set. 
     
     
       19. Apparatus according to  claim 18  wherein the processor is responsive to executing computer instructions to express the vectors as points in the model of noise by projecting each vector onto each Eigenvector thereby representing the original signal as a trajectory in the model of noise. 
     
     
       20. A method according to  claim 11  comprising displaying the time domain signal overlaid with a noise threshold corresponding to the limits of the model of noise, portions of the time domain signal displayed above the noise threshold indicating they are physiologically significant.